Exponentially Moore?

Intel Corporation founder Gordon Moore died on March 24. Moore helped pioneer the development of the silicon chip at the heart of our modern electronic world. But his greatest contribution may have been “Moore’s Law” – his observation in the 1960s that transistor counts on a chip were increasing exponentially, the count doubling every two years.

(Computer scientist Niklaus Wirth pointed out that inefficient software bloats at the same rate, wiping out the gains!)

With Moore’s death, we might ask if Moore’s Law is near death, too? I was in the industry for decades and my trade journals regularly predicted an imminent end to Moore’s Law. Conservative economist Herb Stein pithily said in 1986, “If something cannot go on forever it will stop.”

In order to pack more transistors on a chip, the transistors have to get smaller. Transistors and other chip components are essentially “printed” through a form of lithography on the silicon wafer. Material is deposited and material is etched at different stages. But it involves shining light through a mask, like using a stencil to make a sign. The wavelength of light is about a millionth of a meter (one micron). That was once seen as a limit of how small things could get.

But that limit was passed by using ever shorter wavelengths of “light,” starting with ultraviolet. Moore’s Law has repeatedly dodged death by a series of such innovations. But, surely, there is some hard limit? Current chip features are indeed now just a few atoms across. Atoms would seem to be a hard limit. But, not necessarily.

It may be possible to create multiple layers to chips, expanding from two dimensions to three – thus allowing more transistors on a chip without the transistors getting smaller.

But, such technology reminds me a bit of the world of experimental particle physics. As we probe ever-smaller scales, we need ever-higher energies; meaning ever-larger particle accelerators. In the case of computer chips, the smaller the transistors, the bigger and more expensive the fabrication facility, or “fab.” The cost of building those specialized facilities rises exponentially with each new generation.

A new fab can cost $20 billion. Only a handful of companies can afford that. 90% of the world’s most advanced chips are made in Taiwan. None are made in the U.S. Biden’s CHIPS and Science Act is meant to bring at least some of this business back to the U.S.

What is notable about Moore’s Law is that it actually speaks to a series of revolutionary innovations strung together, each using new technology. What will come next? Quantum computing is one possibility. Another is to use biotechnology to grow computing devices, rather than manufacture them in fabs.

Are there bigger cosmic issues at stake? In 1725, Basile Bouchon invented punched paper tape to semi-automate the monotonous setting up of the day’s silk loom. Moore’s Law can actually be extended back that far, looking at processing power over the decades since.

Ray Kurzweil is a prolific inventor and futurist who believes that Moore’s Law is even more true than Moore realizes. On my first visit to Intel about 20 years ago I watched a video debate between Kurzweil and Moore. Moore said his “law” is not really a law of physics and it will soon run out. But Kurzweil has argued that it can be traced back over the entire history of the universe. Notably, the evolution of living things has been characterized by an exponential rise in complexity of biological information processing over a span of hundreds of millions of years.

Kurzweil argued that if this law has held for so long, it is reasonable to project it a few decades into the future. Doing so would mean the creation of information processing systems that far exceed what our own brains can do. Kurzweil suggests various ways that this might transpire.

It may happen through the printing of ever more transistors (or their equivalent) on a chip. It may happen through genetic alterations of humans. It may involve symbiotic connections of brains and devices. We are already seeing this, for better or worse. Kurzweil acknowledges things could go very wrong. But he also shows a path to a wonderful future that we can be a part of.

Kurzweil refers to “The Singularity” where these systems themselves innovatively design new systems. This could actually exceed current rates of exponential technological growth. In the words of the Eagles’ Don Henley and Glenn Frey, “This could be heaven or this could be hell.” I plan to write more on this in the future.